A New Approach for NOx Soft Sensors for the Aftertreatment of Diesel Engines

To maintain the NOX concentration at an appropriate level, traditionally an air-path control that regulates the intake and exhaust system of diesel engines aims to control the mass air flow and the manifold absolute pressure, which influence the production of NOX. To improve the control accuracy, a more recent approach takes the NOX concentration directly as a controlled output variable, but the sensors monitoring the NOX concentration are slow to respond. Consequently, a direct sensor is inappropriate as a feedback controller. Instead a mechanism called a soft sensor, which computes the NOX concentration from state quantities of diesel engines, is used. Because the prediction accuracy from the sensor model greatly affects the control performance, it is important to improve the model accuracy. However, deviations in the steady state indicate an insufficient model accuracy. This study proposes a method to construct an adaptive NOX soft sensor that corrects the parameters of the sensor model sequentially using the simultaneous perturbation stochastic approximation while comparing the values computed by the software to actual measurements as well as examines the effectiveness of the proposed method experimentally.